| { | |
| "model": "P-JiT 21K", | |
| "architecture": { | |
| "dim": 2944, | |
| "num_heads": 23, | |
| "num_layers": 22, | |
| "ffn_dim": 7936, | |
| "refiner_blocks": 2, | |
| "refiner_ffn_dim": 4096, | |
| "adaln_rank": 256, | |
| "text_dim": 7680, | |
| "text_len": 256, | |
| "in_channels": 3, | |
| "out_channels": 3, | |
| "freq_dim": 256, | |
| "eps": 1e-06, | |
| "bottleneck_dim": 1024, | |
| "patch": [ | |
| 1, | |
| 32, | |
| 32 | |
| ], | |
| "rope_max_positions": 1024, | |
| "rope_spatial_scale": 0.5, | |
| "text_layers": [ | |
| 7, | |
| 15, | |
| 27 | |
| ] | |
| }, | |
| "text_encoder": { | |
| "model": "Qwen3.5-4B", | |
| "extraction_layers": [ | |
| 7, | |
| 15, | |
| 27 | |
| ], | |
| "max_length": 256 | |
| }, | |
| "sampler_defaults": { | |
| "height": 1024, | |
| "width": 1024, | |
| "sampling_steps": 50, | |
| "guidance_scale": 15.0, | |
| "apg_momentum": -0.75, | |
| "apg_eta": 0.0, | |
| "apg_rescale": 10.0, | |
| "noise_scale": 4.0, | |
| "negative_prompt": "", | |
| "noise_sm_count": 132, | |
| "num_timesteps": 1000 | |
| }, | |
| "source": { | |
| "weights": "post-trained", | |
| "num_parameters": 2184091606 | |
| } | |
| } |